Wood anatomy and wood density in shrubs: Responses to varying aridity along transcontinental transects†
The authors thank C. Goedhart, M. Nordenstahl, G. Pongetti, D. Ortega, F. Biganzoli, A. Rolhauser, S. Lambert, M. Mazon, and P. Grierson for help with fieldwork; E. Jobbágy, L. Donovan, B. D. Kloeppel, and J. Ansley for hosting site visits; and N. “Cano” Agüero for permission to conduct research on his land. C. Henry helped with the figures. They also thank three anonymous reviewers and the associate editor for thoughtful comments. Funding for this research was provided by a grant from the Andrew W. Mellon Foundation to H.J.S. and from the National Science Foundation to H.J.S. (award IOS-0641765) and C.S.J. (award IOS-0641569). The scholarship provided by CONACyT to H.I.M.-C. also is appreciated. This research was supported by the National Science Foundation under Cooperative Agreements DEB-9632854 and DEB-0218001, the Coweeta LTER Program. Any opinions, findings, conclusions, or recommendations expressed in the material are those of the authors and do not necessarily reflect the views of the National Science Foundation.
Abstract
Wood density plays a key role in ecological strategies and life history variation in woody plants, but little is known about its anatomical basis in shrubs. We quantified the relationships between wood density, anatomy, and climate in 61 shrub species from eight field sites along latitudinal belts between 31° and 35° in North and South America. Measurements included cell dimensions, transverse areas of each xylem cell type and percentage contact between different cell types and vessels. Wood density was more significantly correlated with precipitation and aridity than with temperature. High wood density was achieved through reductions in cell size and increases in the proportion of wall relative to lumen. Wood density was independent of vessel traits, suggesting that this trait does not impose conduction limitations in shrubs. The proportion of fibers in direct contact with vessels decreased with and was independent of wood density, indicating that the number of fiber-vessel contacts does not explain the previously observed correlation between wood density and implosion resistance. Axial and radial parenchyma each had a significant but opposite association with wood density. Fiber size and wall thickness link wood density, life history, and ecological strategies by controlling the proportion of carbon invested per unit stem volume.
Wood density reflects a plant's carbon investment in its water conducting tissues relative to the size of the stem and, as such, is a key trait correlated with many aspects of a plant's ecology (14), including life history strategy, functional physiology, mechanical properties, and architecture (e.g., 51; 32; 55; 64; 40; 77). Wood density is negatively associated with growth rate (21; 62; 48; 39), but positively associated with survival and life span (48). Dense xylem requires a higher carbon investment for a given increment in stem diameter than light xylem (39, 15). As a consequence, wood density is correlated with successional status: short-lived, fast-growing and -colonizing species have lower densities than long-lived, slow-growing species (67; 71; 48). High intracommunity variation in wood density reported in many studies (e.g., 6; 83; 71) has been explained as the result of niche partitioning (81; 6) and variation in life history strategies (48).
Across broad macro-environmental scales, significant differences in wood density have been correlated with gradients in precipitation (4; 1; 80; 68), temperature (80; 68), and altitude (15; 68). Denser wood has been found to be more resistant to xylem embolisms (30; 32; 36, 38), which may explain why wood density has been observed to increase with aridity.
Wood density is essentially an emergent property, determined by the attributes and types of cells in the xylem. Yet only a handful of studies have quantified the anatomical determinants of wood density across broad environmental gradients (13; 1; 25). Those studies have emphasized trees, which differ from shrubs in biomechanical constraints. Shrubs are relatively understudied compared to trees and yet are highly important growth forms in more arid areas. Previous studies that included mostly shrubs were regionally focused (36, 38) or considered tracheary element properties but not the surrounding matrix (56). The goal of this study was to determine relationships between wood density, xylem anatomy, and aridity in phylogenetically diverse shrub species growing at different aridity levels at various sites along transcontinental transects. The localities studied were situated at the subtropical latitudes of 31–35° in North and South America. They spanned a diversity of ecosystems with similar levels of solar insolation and were minimally impacted by freezing. We chose to focus on varying aridity levels because in addition to its influence on wood density, aridity integrates precipitation and evaporation and is correlated with variation in shrub architecture, hydraulic integration, and the degree of stem segmentation (65).
Wood density appears to be a phylogenetically conserved trait at broad (15; 68) and narrow (12) systematic scales. When phylogenetically conserved traits are correlated with specific environmental conditions, this correlation is frequently interpreted to be the result of environmental filtering (15; 68). The relationship between wood anatomy and density is shaped by the multiple functions of the xylem—storage, mechanical support, and water transport—all of which are interrelated, affected by phylogenetic constraints, and co-vary in response to environmental variables (e.g., water availability; see review in 14). To describe such interactions, we examined correlations between anatomical traits, wood density, and climatic variables and accounted for statistical nonindependence using phylogenetically independent contrasts (23).
Based on a data set of 36 angiosperm species of shrubs and trees, 32 found that resistance to embolism formation was correlated with increasing wood density, which was in turn correlated with increasing vessel wall thickness-to-span ratios and decreasing diameters of the vessels. We sought to confirm this prediction among shrubs alone and to test the hypothesis that both the wall thickness-to-span ratios and decreasing diameters would be correlated with increasing aridity as well as with increasing density. The latter is predicted because narrower vessels tend to have smaller pit areas, which should make them less vulnerable to air seeding (33). Because wood density is strongly affected by abundance and characteristics of fibers (54), we also explore relationships between wood density and fiber vs. vessel traits, questioning the extent to which fiber and vessel traits co-vary independently with wood density.
37 advanced the hypothesis that fibers provide resistance to vessel implosion by buttressing vessels. One possibility arising from this hypothesis is that the proportion of fibers in direct contact with vessels is expected to increase with increasing wood density and increasing aridity. To test this prediction, we determined the percentage of vessel cell wall perimeter that was in contact with fibers in cross section. Moreover, vessel-associated axial and radial parenchyma cells appear to play a role in embolism repair (7; 63, 34). Because embolism repair is thought to be impossible at very low xylem water potentials (18), we tested the hypotheses that the proportions of vessel-associated ray and axial parenchyma cells would increase with decreasing wood density and increasing water availability.
MATERIALS AND METHODS
Study sites
Sixty-one species of shrubs (Appendix S1, see Supplemental Data with the online version of this article) were collected from eight sites in North and South America located between 31° and 35° N or S latitude, respectively (Table 1). At each site, the closest shrub to each of 12–21 randomly chosen points (depending on time constraints and number of researchers at each study site) were sampled so that dominant species had higher chances of being sampled than nondominants. This random sampling protocol resulted in the sampling of four (Coweeta) to 12 species (La Tranca) per site (Appendix S1, see online Supplemental Data). For each site, long-term climate data (30 years), including mean annual temperature (Tmean), minimum monthly temperature (Tmin), maximum monthly temperature (Tmax), and mean annual precipitation (MAP), were obtained from the nearest available weather station (see Table 1). We also include annual precipitation during the two years previous to sampling for comparison to long-term averages; however, only long-term averages were used in the analysis. Mean annual potential evapotranspiration (PET) was estimated from a global database (16; 17) and an aridity index calculated as AI = MAP/PET (74). Note that AI increases with MAP.
| Site | Desert Center | Tucker | Copper Breaks | Whitehall | Coweeta | La Tranca | Cruz de Piedra | El Palmar |
|---|---|---|---|---|---|---|---|---|
| Site coordinates | 33°44′N 115°30′W | 33°43′N 117°37′W | 34°06′N 99°45′W | 33°57′N 83°22′W | 35°00′N 83°30′W | 32°22′S 67°10′W | 33°16′S 66°13′W | 31°53′S 58°14’′W |
| State/Province | California | California | Texas | Georgia | North Carolina | San Luis | San Luis | Entre Rios |
| Country | USA | USA | USA | USA | USA | Argentina | Argentina | Argentina |
| Date sampled | January, May 2005 | May, June 2004 | August 2005 | July 2004 | July 2004 | January, March 2005 | January, March 2005 | January 2005 |
| Weather station | Hayfield Pumping Plant | Tustin Irvine Ranch | Copper Breaks SP | Athens, Ben Epps Airport | Coweeta Exp. Station | El Retamo (Mendoza) | San Luis Aero. | Gualeguaychú and Concordia |
| Source of climate data | NCDC, COOP | NCDC, COOP | NCDC, COOP | NCDC, COOP | NCDC, COOP | 22 | SMNA | SMNA |
| Station coordinates | 33°42′N, 115°38′W | 33°42′N, 117°45′W | 34°07′N, 99°45′W | 33°57′N, 83°20′W | 35°04′N, 83°26′W | 32°27′S, 67°24′W | 33°16′S, 66°21′W | Concordia: 31°18′S, 58°01′W Gualeguaychú: 33°00′S, 58°37′W |
| Distance, station to site (km) | 8 | 9 | 2 | 5 | 1 | 23 | 13 | Concordia: 68; Gualeguaychú: 122 |
| MAP (mm) | 100 | 325 | 663 | 1250 | 1850 | 250 | 680 a | 1190 |
| AP2 prev (mm) | 16/223 | 164/293 | 331/877 | 1176/1272 | 1760/2132 | low/high b | low/high b | high/low b |
| PET (mm) | 1585 | 1143 | 1450 | 1018 | 900 | 1141 | 1149 | 981 |
| Tmean (°C) | 20.8 | 17.3 | 17.2 | 16.7 | 12.7 | 19.8 | 17.2 | 18.4 |
| Tmin (°C) | 3.4 | 4.7 | −3.0 | 1.4 | −3.8 | 3.2 | 3.4 | 6.7 |
| Tmax (°C) | 40.1 | 29.7 | 37.7 | 31.5 | 28.7 | 33.4 | 31.2 | 32.5 |
| Aridity | 0.06 | 0.28 | 0.46 | 1.23 | 2.06 | 0.22 | 0.59 | 1.21 |
| Seasonality | W | W | Sp | N | N | S | S | S |
| Vegetation | Desert scrub | Sage scrub | Mesquite savanna | Hardwood forest | Hardwood forest | Desert scrub | Mesquite savanna | Palm forest |
| Mean wood density (SD) | 0.79 (0.24) | 0.69 (0.08) | 0.76 (0.06) | 0.59 (0.07) | 0.57 (0.05) | 0.78 (0.11) | 0.61 (0.13) | 0.53 (0.11) |
- a Notes: AI = aridity index MAP/PET, AP2 prev = annual precipitation during the two years previous to sampling, EF = experimental forest, LTER = long-term ecological research site, MAP = mean annual precipitation, NCDC = National Climatic Data Center, NP = national park, PET = mean annual potential evapotranspiration, S = Seasonality of precipitation (N = no pronounced seasonality, S = summer maximum, Sp = spring maximum, W = winter maximum), SMNA = Servicio Meteorológica Nacional, Argentina; SP = state park, Tmax = maximum monthly temperature, Tmean = mean annual temperature, Tmin = minimum monthly temperature, WS = wildlife sanctuary.
- a Corrected for elevation using the climate model by 35.
- b Estimated based on data from the weather station listed. Accurate data for these years are not available for these sites.
This study was designed as a correlation analysis, with each individual plant sampled representing a single data point for each measured anatomical trait to be analyzed with respect to variation in aridity and wood density. The study was not designed to characterize differences between species. To test for the probability that the individuals sampled might represent outlying values for their populations, we analyzed additional individual samples for four species. Among all variables of three species (one species is not considered because we had only two additional samples), 90% of the values of the individuals used in the correlation analysis were less than 20% different from the mean of a different set of individuals; average differences across all values of anatomical characters and wood density was ±10.27% (Appendix S2, see Supplemental Data with the online version of this article). However, we recognize that inclusion of many species at the expense of reduced replication within species means that individual measurements of density and anatomy may not be representative for population or species mean trait values.
Wood density and anatomical measurements
Wood samples were taken from main stems near ground level and were stored in the laboratory in air-dried condition until wood density determination. Three small wood samples (≤1 cm3) from the outermost growth rings of each species were rehydrated for at least 12 h under partial vacuum until they no longer floated. Sapwood density was analyzed as described by 31 using the Archimedes principle: wood samples were immersed in a beaker of water that was placed on a high-precision balance, and the weight of water displacement was converted into volume using an equation that adjusts for changes in the density of water with water temperature. Afterward, samples were placed in a drying oven at 75°C for 48 h and then weighed immediately after cooling to obtain dry mass.
Anatomical traits were measured on the outermost 1–2 growth rings of the same basal stem samples used to determine density. Samples were softened in boiling water for 2 h, then sectioned with a sliding microtome. Thin sections (20–30 µm) were dehydrated using an ethanol dehydration series through 100% ethanol, transferred to xylene, and mounted with Permount (Biomeda, Burlingame, California, USA). In most cases, xylem sections were stained with both safranin and fast green to distinguish lignified cell walls from cytoplasm.
To calculate the area occupied by each cell type (as percentage of total image area) and cell dimensions, we drew the lumens and walls of approximately 200 vessels and fibers per sample in randomly selected radial sectors of the two most recent growth rings using a graphics tablet (Intuous 3, Wacom, Kita Saitama-Gun, Saitama, Japan) (Fig. 1). The area occupied by each anatomical trait was calculated from the drawings using XTools in the program ArcView (version 3.2, ESRI, Redlands, California, USA). Imperforate tracheary elements, i.e., cells derived from fusiform cambial initials that have secondary walls but do not have end wall perforations, (10), such as libriform fibers, fiber-tracheids, and small-diameter tracheids, were lumped into a category hereafter referred to as fibers (e.g., 37, 36). Vessels and fibers were drawn until the desired sample size was reached, in most cases in more than one radial sector. The areas of each cell type, including axial and radial parenchyma, were determined from the same sectors. Vessel lumen diameters were calculated as the diameters of circles with the same area of individual vessel lumens (41) based on the shape files (drawings) described. The same method was used to measure fiber lumen diameters. Vessel lumen diameter, including the mean (dmean), maximum (dmax, a surrogate of theoretical conductivity; 11), and hydraulic mean (dh, a measure of the hydraulic contribution of conduit diameter; 41), were calculated as diameters of circles that had the same areas as the individual vessel lumens (41). The same method was used to measure fiber lumen diameter. Vessel hydraulic mean (dh) was calculated as the sum of the contribution of all conduit diameters (∑d5) divided by the total number of vessels (∑d4) (41; 19).
Example of the Environmental Systems Research Institute (ESRI) shape files analyzed with the program ArcView. Drawing of a cross section of Ambrosia dumosa. Light blue = ray, dark blue = fiber lumen, white = fiber wall, pink = axial parenchyma, gray = vessel wall, yellow = vessel lumen.
To estimate the safety factor for vessel implosion, (t/b)h2, we measured the total thickness of walls of adjacent vessels, t, and the diameter of the conduit closest to the mean hydraulic diameter, b (32). The safety factor for vessel implosion was calculated for an average of 25 vessel pairs per species in which at least one of member of a vessel pair was within ± 10 µm of the hydraulic mean diameter.
For the anatomical analysis quantifying the types and number of cells that are the nearest neighbors of vessels, 5–10 vessels were randomly chosen in the same wedge section used to measure total area and cell dimensions. Per vessel, we determined the number of different cell types in direct contact with the vessel and the percentage of the vessel perimeter in contact with the different cell types (i.e., the length of the arc of various cell types per vessel perimeter).
Data analysis
Anatomical traits and wood density were analyzed for their correlations with MAP, PET, AI, Tmean, Tmin, and Tmax. Correlations among anatomical traits and of the traits with wood density were also calculated. To detect broad trends in variation among multiple anatomical traits and wood density, we performed a principal component analysis (PCA). Because anatomical data were only available from three South American sites, the statistical power was lacking for a comparison between the two hemispheres, and all sites were therefore analyzed together. Correlations and PCA were carried out using the program Statistica (version 6, StatSoft, Tulsa, Oklahoma, USA). To examine the effects of statistical nonindependence resulting from shared phylogenetic histories, we calculated phylogenetically independent contrasts correlations (PICs) using the module PDAP:PDTREE (45) in the program Mesquite (version 2.5; 43). Phylogenetic relationships among species (Appendix S3, see Supplemental Data with the online version of this article) were reconstructed using the program PHYLOMATIC (79) and were refined using published phylogenies (52; 75, 76; 42; 24; 29; 82; 44). Polytomies were treated as soft, and the degrees of freedom were reduced (57; 26). Because Eupatorium buniifolium Hook. & Arn. was the only species present in two different sites, the samples were analyzed as separate taxa. Characters that were not adequately standardized (i.e., if the absolute values of the standardized PICs and their standard deviations showed a significant association) (28, 27) were log, arcsine, or square-root transformed depending on the nature of the data. All branch lengths were set to one for the independent contrast analysis because the tree is based on multiple data sets. We recognize that the PIC correlations presented here are based on trait values that may not be representative of species means because of the small sample sizes, and therefore our ability to infer from tests of statistical nonindependence in this data set to correlated evolution across species is limited. For all correlations, statistical significance was determined in two-tailed tests after correcting for false discovery rate in multiple comparisons using the Benjamini and Hochberg procedure (5; 78).
RESULTS
Broad trends
First, second, and third PCA axes explained 31%, 18% and 12% of the variation, respectively. In the first principal component axis, axial parenchyma area, wood density, and climate variables had high loadings, as did all fiber traits except fiber wall area and fiber wall thickness. Vessel traits, the vessel implosion resistance parameter, and fiber wall thickness were important in the second principal component, while cell areas (except fiber lumen and parenchyma areas) had high loadings in the third principal component (data not shown). The PCA graph (Fig. 2) summarizes the general trends observed in the correlation analyses (Appendices S4 and S5, see online Supplemental Data). Variation in fiber and vessel traits was orthogonal. Wood density varied independently from vessels traits but was strongly related to fiber variation. Climate variables were more intimately associated with fibers than with vessel traits.
Loadings plot of anatomical traits and climate variables on the first two principal components axes. tpa = total parenchyma area, vla = vessel lumen area, vwa = vessel wall area, MAP = mean annual precipitation, PET = potential evapotranspiration, AI = aridity index, Tmean = mean annual temperature, Tmin = minimum monthly temperature, Tmax = maximum monthly temperature, (t/b)h2 = safety factor for vessel implosion.
Wood density and anatomical traits in relation to climate
Wood densities observed in this study ranged from a minimum of 0.41 g⋅cm−3 for Daphnopsis racemosa Griseb. (Thymelaeaceae) to a maximum of 1.11 g⋅cm−3 for Ambrosia dumosa (A. Gray) Payne (Asteraceae). Wood density was negatively correlated with MAP and AI, positively correlated with PET Tmax and Tmean and was not correlated with Tmin. (Table 2). Only significant correlations with MAP and AI persisted when PICs were used.
| Climate variable | r | PIC r |
|---|---|---|
| MAP (mm−3) | −0.53*** | −0.44** |
| AI | −0.53*** | −0.44** |
| PET (mm−3) | 0.47** | 0.29 |
| Tmax | 0.39** | 0.13 |
| Tmean | 0.36* | 0.16 |
| Tmin | −0.15 | −0.05 |
- a Notes: * P < 0.05, ** P < 0.01, *** P < 0.001. r = phylogenetic uniformed correlation coefficients (N = 62), PIC r = phylogenetic independent contrast correlation coefficients (N = 61).
The anatomical traits significantly associated with most climate variables were fiber traits (Fig. 2; for specific trait correlations, see online Appendix S4) with the exception of fiber wall thickness. Fiber traits showed stronger correlations with AI and MAP than with temperature variables (Fig. 2). Fiber lumen traits decreased and fiber wall/lumen ratio and fiber wall areas increased with decreasing MAP and AI. No vessel traits were associated with climate variables. The safety factor for implosion (t/b)h2, a trait strongly correlated with wood density, was, as hypothesized, also correlated negatively with MAP and AI and positively with PET but not temperature. An unexpected finding was that ray area was significantly positively correlated with MAP and axial parenchyma was negatively correlated with MAP, AI, and positively with Tmax and Tmean. In general, associations between anatomical traits and climate were weaker in correlations using PICs. Relationships between parenchyma and MAP and Tmax temperature did not persist using PICs. In two cases, correlations were only significant using PICs, i.e., vessel wall area was negatively correlated with AI and MAP (for details, see online Appendix S4).
Anatomical traits in relation to wood density
Among anatomical traits, the PCA discerned two distinct axes of variation: fiber traits and wood density were independent from the axis describing variation in vessel traits, which were not correlated with wood density (Fig. 2; for details on anatomical trait correlations, see online Appendix S5). Wood density was positively correlated with fiber wall to lumen ratio (Fig. 3A) and fiber wall area (Fig. 4) and negatively correlated with fiber lumen diameter (Fig. 3B) and fiber lumen area (Fig. 4). Fiber wall thickness was negatively correlated with wood density because fibers with wide lumens had thicker walls (in absolute terms), and wide-lumened fibers prevailed in low density woods (Fig. 3B).
Correlations between wood density and (A) log of fiber wall to lumen ratio, (B) log fiber lumen diameter, (C) ray area and (D) parenchyma area. * P < 0.05, ** P < 0.01, *** P < 0.001. Insets represent phylogenetic independent contrast correlations; the regression lines were forced through the origin.
Proportional change in mean cross-sectional areas of wood anatomical traits as a function of wood density.
Total parenchyma area varied independently from wood density, but when divided into radial and axial parenchyma, each type of parenchyma had a significant and opposite relationship with wood density (Fig. 3C and 3D, respectively; Fig. 4). Axial parenchyma area increased strongly with wood density up to about 0.8 g⋅cm−3 (Fig. 3D). Ray area decreased with increasing wood density using both raw and independently contrasted trait values (Fig. 3C)
Relationships among anatomical traits
With the exception of fiber wall thickness, variations in fiber and vessel traits were orthogonal (Fig. 2). Fiber traits, along with parenchyma area, had high loadings on the first principal component. Fiber diameter was positively associated with fiber lumen area and negatively correlated with fiber wall area and wall to lumen ratios (online Appendix S5).
Theoretical resistance to implosion, (t/b)h2, was positively correlated with fiber wall area and fiber wall to lumen ratio and negatively correlated with vessel diameter (Fig. 5A) and fiber lumen diameter. Axial parenchyma area increased as both fiber lumen diameter and total fiber lumen area decreased (Fig. 5B) and was strongly positively correlated with fiber wall/lumen ratios. Ray parenchyma area tended to have the opposite trend and was positively correlated with fiber lumen diameter and negatively correlated with fiber wall area (Fig. 5C). Most of the intertrait correlations remained significant after correcting for statistical nonindependence. (online Appendix S5).
Relationship between selected anatomical traits. (A) Log mean vessel diameter and log theoretical resistance to implosion, (B) axial parenchyma area and fiber lumen area, and (C) ray area and fiber wall area, (D) log fiber wall thickness and log mean vessel diameter. ** P < 0.01, *** P < 0.001. Insets represent phylogenetic independent contrast correlations. Regression lines in the insets were forced through the origin.
The second principal component reflects primarily vessel lumen diameter, vessel density, and fiber wall thickness (Fig. 2). These trends resulted from a strong negative interaction between vessel size and number (Appendix S5) and a positive correlation between fiber wall thickness and mean vessel diameter (Fig. 5D) and vessel density. In essence, fibers with larger lumens also had thicker cell walls in absolute terms, but overall had much lower fiber wall thickness/\lumen diameter ratios.
Nearest neighbors to vessels
The total number of cells immediately adjacent to vessels was positively associated with wood density in the nearest neighbor analysis (Table 3), in part because in denser xylem, cells in the matrix surrounding vessels were smaller and more closely packed. Thus, wood of plants from dry places (low AI, low MAP, and high PET, temperature) had more cells surrounding vessels. This correlation with cell number in contact with vessels and climatic variables did not hold when PICs were used.
| ρ (g⋅cm−3) | MAP (mm−3) | PET | AI | Tmax | Tmean | Tmin | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Vessel nearest neighbors | r | PIC r | r | PIC r | r | PIC r | r | PIC r | r | PIC r | r | PIC r | r | PIC r |
| Total number of cells | 0.41** | 0.08 | −0.45** | −0.28 | 0.34* | 0.14 | −0.43** | −0.24 | 0.32 | 0.12 | 0.41** | 0.26 | 0.007 | 0.08 |
| Axial parenchyma | 0.24 | −0.01 | −0.48** | −0.24 | 0.44** | 0.24 | −0.49** | −0.25 | 0.45** | 0.27 | 0.47** | 0.23 | 0.12 | −0.01 |
| Fibers | 0.04 | 0.09 | 0.36* | 0.15 | −0.35* | −0.21 | 0.39** | −0.17 | −0.35* | −0.23 | −0.38** | −0.16 | −0.19 | −0.03 |
| Vessels | −0.28 | −0.08 | −0.08 | −0.03 | 0.12 | 0.14 | −0.11 | −0.06 | −0.01 | −0.01 | 0.02 | −0.1 | 0.09 | 0.15 |
| Ray | −0.22 | −0.06 | 0.38** | 0.21 | −0.34* | −0.19 | 0.37* | 0.22 | -0.025 | −0.06 | −0.23 | −0.16 | 0.02 | 0.2 |
- a Correlation coefficients were calculated using the percentage of the vessel perimeter in contact with each type of cell, except for total number of cells which represents the mean number of cells in direct contact with vessels. * P < 0.05, ** P < 0.01, *** P < 0.001. r = phylogenetic uniformed correlation coefficients, PIC r = phylogenetic independent contrast correlation coefficients.
We examined the proportion of cells adjacent to vessels that were fibers using counts as well as the percentage of the circumference of vessels in direct contact with fibers to account for possible effects of differences in fiber size. Our hypothesis that the proportion of fibers directly associated with vessels might increase with both increasing wood density and increasing aridity was not supported (Table 3). The number of fibers in contact with vessels was not significantly correlated with wood density (data not shown); however, the percentage of fiber wall in contact with vessels was positively associated with AI (i.e., increasing precipitation). As a result of their large size, fibers surround a greater percentage of vessel perimeters in sites with lower PET. Vasicentric tracheids were observed in 23% of the species studied, exclusively in shrubs from water-limited environments with MAP <700 mm.
All climate variables except Tmin had significant relationships with the proportion of axial parenchyma in direct contact with vessels. Axial parenchyma increased with PET, Tmax, and Tmean and decreased with MAP and AI. More ray cells were in contact with vessels as precipitation increased and AI and PET decreased (Table 3). However, none of these associations were significant with PICs.
DISCUSSION
Wood density variation across the aridity levels
Wood density determinations in this study resulted in a few surprisingly high values, for example, for the desert shrubs Ambrosia dumosa and Larrea tridentata (Appendix S1, see online Supplemental Data). Overall, our data did not differ significantly from those reported in the literature (analysis not shown), but a tendency toward higher densities may be explained by the fact that our samples were taken from the base of the shrubs, while previous data are mostly for small branches. In shrubs, branches are more expendable than the stem base (53), and branches usually have lower wood densities than stems (69).
In contrast to 68, who found strong correlations with temperature variables in a large database of woody species (predominantly trees), this study of shrubs from environments with a narrower temperature range found stronger associations of wood density with MAP and AI. In fact, these were the only climate variables that remained significantly correlated with wood density when phylogeny was considered.
Wood density is driven by fiber traits
Increases in wood density among shrubs were driven largely by decreasing fiber lumen diameters, higher fiber wall to lumen ratios, and total wall areas. In essence, wood density was driven by reduction in fiber cell sizes, an observation that agrees with previous findings for chaparral shrubs (37, 38). In our study, fiber wall thickness increased with lumen diameter, so wall thickness alone was negatively correlated with wood density, a finding that contrasts with previous studies of hardwood trees (1; 25). Dense woods have more fiber cells per unit area than do light woods. Regardless of the abundance and characteristics of other cell types, individual species can achieve particular densities by varying fiber lumen diameters and wall areas. High wood density can occur even when the percentage of thin-walled cells (i.e., parenchyma or rays) is elevated.
A positive correlation between cavitation resistance and fiber wall area led 37 to suggest that the fiber matrix reinforces vessel walls. A corollary prediction is that dense woods might be characterized by proportionately more buttressing fibers in direct contact with vessels (Fig. 6 in 37). We did not find support for this prediction: the proportion of fiber wall in contact with vessels did not increase with wood density and decreased with increasing AI and MAP.
Vessel traits are independent of wood density in shrubs across broad aridity levels
A direct spatial tradeoff between cell wall material and vessel lumen area would predict that wood density should be inversely related to vessel lumen area (56). Indeed, previous studies have reported negative relationships between wood density and total vessel area (36) or vessel area and vessel fraction (56). We found, however, that across the 61 shrub species in this study, total vessel lumen areas and vessel diameters were independent of wood density. There was no evidence for a direct spatial tradeoff between wood density and conduction capacity, suggesting that the density of the matrix surrounding vessels regulates wood density independently of changes in other cell types. In contrast, some studies of wood density variation within tree species or within individual trees found significant relationships between wood density and vessel traits (60; 61; 72, 73; 49, 47). More research is required to determine whether wood anatomies of shrubs and trees reflect fundamental differences in biomechanics, hydraulics, and pathway lengths over which water has to be transported.
Fiber traits, vessel implosion, and wood density
Theoretical implosion resistance (t/b)h2 and fiber traits are strongly linked to wood density, but surprisingly, the link between fiber traits and (t/b)h2 is weak and only present among raw trait values. These findings do not throw much light on the previously reported the relationship between wood density and (t/b)h2 (30; 37, 36, 38). 32 and 37 suggested that the theoretical implosion resistance of vessels (t/b)h2 may be associated with stronger mechanical support from fibers. This study does not support a role of fibers as direct mechanical support for vessels because (t/b)h2 was not correlated with the proportion of fibers in contact with vessels (data not shown). The observation that there are vessels in plants in very dry environments that are disproportionately resistant for their wood density (36) indicates that high (t/b)h2 is of great selective advantage under drought. Interestingly, in this study, PET showed a high correlation with vessel resistance to implosion, indicating that the evaporative demand is important in the variation of this trait. However, the role of fibers, and thus of wood density, in determining implosion resistance is still an unanswered question and leaves open the potential for additional functions of the fiber matrix in drought resistance, e.g., in hydraulically isolating vessels (9; 85; 65).
Ray and axial parenchyma show opposite associations with wood density
Opposite trends in correlations between axial and radial parenchyma with wood density suggest different functional dimensions for each type of parenchyma cells. The lack of significant correlation between axial parenchyma area and wood density using independent contrasts shows strong phylogenetic influence on the relationship.
Xylem parenchyma is a relatively thin-walled tissue, which means that wood that is rich in parenchyma may need additional mechanical support from fibers. The negative association between axial and total parenchyma areas and fiber lumen area may be an indication that large parenchyma areas can compromise mechanical stability when combined with extensive fiber lumen space. Alternatively, the interaction might reflect a coordinated strategy that mediates two different axes of variation: plants in moderately dry environments increase wood density by reducing fiber lumen while increasing storage capacity by increasing axial parenchyma.
Xylem parenchyma is important in water transport in providing a temporary reservoir for water to prevent embolism formation in vessels and in providing water and osmotic agents for embolism repair (8; 3; 7; 2; 63; 20). Moreover, rays are the radial pathway for flow of Münch-water between phloem and xylem (46; 34). The negative correlation between ray area and density in shrubs contradicts reports that the amount of ray tissue in trees is positively associated with wood density (70; 59). The difference may be due to the fact that shrubs occur in much drier environments than trees.
Large cross-sectional areas of stems occupied by rays in environments where wood density is high may be a disadvantage if there are large temporal gaps in water supply. Because the water used to prevent embolism formation or to refill embolized vessels comes from other cells within the stem, the plant would be investing water potentially used in other functions (e.g., phloem transport) to restore transpiration. This water will be lost into the atmosphere. In environments where water supply is replenished regularly (daily cycles), this “water loan” will be restored to the lender cells or tissues. However, if limited water availability persists, repair would cause water from other tissues, including Münch-water, to be lost through transpiration. Less-developed radial systems or less connectivity between phloem and xylem in arid sites could restrict the negative effects of embolism prevention or repair. Further indirect support for this hypothesis is provided by plant species from the driest sites that have relatively large rays but practically no contact between rays and vessels, such as Ambrosia dumosa and Hymenoclea salsola A. Gray in the Californian desert or Prosopis alpataco Phil. and Senna aphylla (Cav.) H. S. Irwin & Barneby in the Argentinean desert.
In summary, our results suggest that variation in wood density is mainly driven by variation in fiber lumen diameter, which is directly related to cell size and to cell wall thickness. Interestingly, the simple structural characteristic of fiber diameter potentially integrates wood density and life history strategies in a manner similar to the structural controls thought to account for the worldwide leaf economic spectrum (84), i.e., the size–number relation of living cells within the leaves, (58; 50) and the variation in cell wall thickness (66). The general importance of structural tradeoffs between cell size and proportional cell wall thickness to ecological strategy merits further investigation at multiple levels of structural organization in plants.
